Loudness Integration

[splrk]

Are there any plans to incorporate loudness into the Rivendell import system?

[itsalexjones]

+1 for this, we're currently manually normalising before import, but a loudness normalisation option would be excellent

[elmuz]

I'd love this feature too, but I'd move normalization phase during playout and not in importing. My idea for this is the following: 1) We should keep all ingested material in a 1:1 copy of original source, without applying any destructive filter/gain. While importing two values for each song would be stored in database: the loudness (global, gated, in LUFS) and the true peak value. 2) In RDAdmin -> Settings (if we want to keep it global) or in Hosts settings (if we set per host) we configure a reference playout loudness level, i.e. -23 LUFS 3) RDAirPlay when playing each track would apply a gain on the fly according the level of loudness of each track and the reference set in RDAdmin. However it would raise (if needed) the song up to an undistorted level (looking up the true peak of the track).

[itsalexjones]

I know that RCS Zetta does it this way, and works very well. Might also be useful to be able to set a different loudness target for different groups? For instance, we normalise our imaging to -19 and our songs to -23LUFS which helps make sure the imaging doesn't get lost. I know this can also be done in Zetta.

[danielbair]

I took someone else's code for ebur128 loudness measurement and the rivendell library and edited the code to create a simple script called rdnormalize about three years ago. You can find it here... https://sourceforge.net/projects/rrakstudio/files/DEBs/

[ermina]

External scripts that scan loudness (using https://github.com/jiixyj/libebur128) of CUTS and then edit the PLAY_GAIN field work fine, but an option to use loudness normalization instead of peak normalization where Rivendell proposes it would be great. The "load from file" feature in RDCartslots is a prime example of where peak normalization is wrong. Loading a modern music mp3 (with 2dB dynamic range) next to an interview straight out of the field recorder (with loud peaks but low RMS, as any voice recording is) is a common use case in real life that would greatly benefit from the availability of loudness alignment.

[tklicki]

+1 for integrating this feature natively

[SM0VXI]

+1 for native implementation.